HQ Player

[Account Closed]

Jussi,

 

Do you expect that these new modulators will run on the Windows version of HQP? Have you been testing them on a Win10/64 system yet? From what I have been reading above, it looks questionable. I hope they do as I would love to try them.

[Le Concombre Masqué]

10 hours ago, Miska said:

 

Something like this:

https://www.supermicro.com/en/products/motherboard/X11DAi-N

https://www.asus.com/Motherboards/WS-C621E-SAGE/

 

 

HQPlayer OS doesn't support CUDA, but the package built for Ubuntu Server 18.04 LTS does support CUDA. It would be still quite different from running something big and heavy as OS.

 

So that would make a D4 iteration closer to Embedded's SQ?

 

for, commercially, the package built for Ubuntu Server 18.04 LTS  is a Desktop 4 purchase, upgrade afaic, correct  ? Then I would just have to erase my Bootcamp partition and follow such recipe : https://www.lifewire.com/dual-boot-linux-and-mac-os-4125733 ? And then later simply do that on a more powerful machine ? New, for purchasing a GPU extension would not be a viable solution any longer due to the need of a as powerful processor as possible for EC modulators, correct?

 

Since Audirvana works as front end, when I can borrow the MBP that has it, I find Embedded so handy dandy with  just to have to press option at boot to select your images on usb stick... Would it be possible that D4  be recognised by Audirvana just as Embedded, as an alternative to Client?

 

[Miska]

48 minutes ago, Le Concombre Masqué said:

for, commercially, the package built for Ubuntu Server 18.04 LTS  is a Desktop 4 purchase, upgrade afaic, correct  ?

 

No, Embedded is available as HQPlayer OS image and as a package for following Linux distributions:

• Ubuntu Server LTS (x64 arch + CUDA)
• Debian Stable (x64 and arm64 arch)
• Fedora Minimal latest release (x64 arch)

Embedded license key doesn't care about the used OS variant, only about the hardware fingerprint.

 

53 minutes ago, Le Concombre Masqué said:

Then I would just have to erase my Bootcamp partition and follow such recipe : https://www.lifewire.com/dual-boot-linux-and-mac-os-4125733 ? And then later simply do that on a more powerful machine ? New, for purchasing a GPU extension would not be a viable solution any longer due to the need of a as powerful processor as possible for EC modulators, correct?

 

I really cannot say much about running these things on a Mac hardware, only about regular PCs... And I think on more recent Macs with T2 security chip you won't be able to run any of this anymore. Apple allows only macOS and Windows.

 

56 minutes ago, Le Concombre Masqué said:

Would it be possible that D4  be recognised by Audirvana just as Embedded

 

No, unless Damien adds support for the HQPlayer control protocol (like Roon does for example).

 

[Miska]

5 hours ago, bobflood said:

Do you expect that these new modulators will run on the Windows version of HQP? Have you been testing them on a Win10/64 system yet? From what I have been reading above, it looks questionable. I hope they do as I would love to try them.

 

I have not tested yet, but I would expect it to work...

[Miska]

On my upgraded server, i7-8086K does RedBook to DSD256 using poly-sinc-ext2 and ASDM7EC just fine:

 

DSD512 doesn't work with the ASDM7EC as it is now, unless I find some more optimizations...

 

 

[luisma]

I have to try the new modulators, DSD5, ASDM5 or ASDM7, where you guys find the bigger difference? I will have to go most likely to DSD128

Edit: And yes it depends on the filter used too I guess I will be using poli mqa mp

 

On PCM is the RPDF dither new?

 

[Le Concombre Masqué]

8 hours ago, Miska said:

 

No, Embedded is available as HQPlayer OS image and as a package for following Linux distributions:

• Ubuntu Server LTS (x64 arch + CUDA)
• Debian Stable (x64 and arm64 arch)
• Fedora Minimal latest release (x64 arch)

Embedded license key doesn't care about the used OS variant, only about the hardware fingerprint.

 

 

I really cannot say much about running these things on a Mac hardware, only about regular PCs... And I think on more recent Macs with T2 security chip you won't be able to run any of this anymore. Apple allows only macOS and Windows.

 

 

No, unless Damien adds support for the HQPlayer control protocol (like Roon does for example).

 

Maybe my future machine could be a Mac, still : https://www.omgubuntu.co.uk/2018/11/apple-t2-chip-cant-boot-linux

 

Anyway, it's good to know that starting the use of Embedded  on my old MBP, it will be a transferable license of the same product  on the future machine 

[luisma]

Listening to the new modulators now, can only do DSD128 with 7EC but I am pleasantly surprised, more by the fact that Jussi knowing how improved the modulators are haven't praised these too much publicly leaving the feedback to the community. 

I was exclusively doing straight PCM, not because HQPlayer is bad but because with my specific system (dac speakers etc) PCM is very detailed which I like, now I hate to have to accept this but software algorhtym alone may have improved over the dacs transistor ladder. I'm testing poli mqa mp 1x and nx, I can only imagine how other filters and DSD512 may sound. The modulator clarity and separation and noise it's amazing 

 

Last but certainly not least the algorhtym is someone's creation after much experience in this field, very well done sir, IMO this makes more of difference than filters when fed with high quality source content. 

 

[luisma]

I would love to read a review or comparison of a good system with AL and HQP+EC mods vs million taps device. 

 

BTW EC? Error Correction? Extended compensation? 

[jimdukey]

I missed a few posts here.

What new Modulators?

I'm on 4.04 Desktop, and see the same list I've seen for a long while.

Ive been using ASDM5, seen it recommended here for DSD 128.

I seemed to like it more than ASDM7.

I've noticed that some who get DSD 256 or 512 don't use those Modulators,

but the ones not indicated for those rates instead.

 

[elan120]

7 minutes ago, jimdukey said:

What new Modulators?

I'm on 4.04 Desktop, and see the same list I've seen for a long while.

No new modulators for desktop version, the three new modulators are for Embedded that was released on 7/4/2019.

[luisma]

My bad, I should have posted on embedded

[dminches]

Upgrade to 4.x Questions

 

I am currently running HQP 3.X with Roon.  When I upgrade to 4.x what should I run, server or client and should I only be installing the one I need?

 

Also, I am running this in WS 2016 with audiophile optimizer.  Are there any compatibility issues with WS 2016?

 

[Miska]

1 hour ago, luisma said:

I would love to read a review or comparison of a good system with AL and HQP+EC mods vs million taps device. 

 

BTW EC? Error Correction? Extended compensation? 

 

You can have million taps (or 16 million) and the new modulators, both! The "megatap" filters have been in HQPlayer for quite a while already. 

 

Extended Compensation, didn't want to start writing a book in the modulator name, so needed to come up with something short...

 

1 hour ago, jimdukey said:

I'm on 4.04 Desktop, and see the same list I've seen for a long while.

 

Not yet there, but coming to 4.1.0...

 

[Miska]

2 minutes ago, dminches said:

I am currently running HQP 3.X with Roon.  When I upgrade to 4.x what should I run, server or client and should I only be installing the one I need?

 

You need at least server, client is optional when you use Roon. But may make life easier in some cases still.

 

3 minutes ago, dminches said:

Also, I am running this in WS 2016 with audiophile optimizer.  Are there any compatibility issues with WS 2016?

 

Server versions of Windows have never been officially supported or recommended. It may or may not work, but I don't expect any big differences to v3.

 

[luisma]

43 minutes ago, Miska said:

You can have million taps (or 16 million) and the new modulators, both! The "megatap" filters have been in HQPlayer for quite a while already. 

 

I was being ironic at..... You know, the meaningless of taps in some claims. 🙂

 

[k6davis]

17 hours ago, Miska said:

On my upgraded server, i7-8086K does RedBook to DSD256 using poly-sinc-ext2 and ASDM7EC just fine. DSD512 doesn't work with the ASDM7EC as it is now, unless I find some more optimizations...

 

My current server, with an AMD Ryzen 7 2700x, can easily handle upsampling all formats to DSD5EC DSD128. I'm hoping that the new AMD chips (or something else that's not too outrageously priced) can get us to EC at DSD512, at least with RedBook. Then it'll be possible to use the 1x (EC) and Nx (non EC if necessary) settings to automatically play all of the music at DSD512 with the best possible sound quality.

 

5 hours ago, luisma said:

Listening to the new modulators now, can only do DSD128 with 7EC but I am pleasantly surprised, more by the fact that Jussi knowing how improved the modulators are haven't praised these too much publicly leaving the feedback to the community. 

I was exclusively doing straight PCM, not because HQPlayer is bad but because with my specific system (dac speakers etc) PCM is very detailed which I like, now I hate to have to accept this but software algorhtym alone may have improved over the dacs transistor ladder. I'm testing poli mqa mp 1x and nx, I can only imagine how other filters and DSD512 may sound. The modulator clarity and separation and noise it's amazing 

 

Last but certainly not least the algorhtym is someone's creation after much experience in this field, very well done sir, IMO this makes more of difference than filters when fed with high quality source content. 

 

 

I couldn't agree more. These new EC modulators are fantastic. There are aspects of the sound that are lost going from DSD512 to DSD128, but there are also aspects that are gained from the EC modulators (available at the lower bitrate). Forced to choose, for now, I'll take the lower bit rate. EC is that good. 

[ted_b]

Jussi, I look forward to finally getting my year (and almost a half) long hiatus (not volunteered; move to Colorado-related) over and get back to my system.  When I do I expect to be doing the same upsampling with my Holo Spring...that is, PCM to 352/384 and all DSD to DSD512.  However, with the introduction of the new EC modulators I may be stymied with DSD512 until I update my up-til-now heavy lifting I7-6700k.     Question: does the fact that I don't do PCM to DSD give my current server any chance to do DSD512 with the ECs, or is DSD64 to DSD512 still too much data?

[rando]

From the previous page in reply to why DXD is more intensive than RBCD to upsample into DSDxx

 

On 7/6/2019 at 7:05 AM, Miska said:

Because there is more data to process.

 

A few posts later...

 

On 7/6/2019 at 4:22 PM, Miska said:

Generally, regardless of modulators, I would say;

DSD64 -> DSD128 is a huge step

DSD128 -> DSD256 is still a big step

DSD256 -> DSD512 is medium to small step

DSD512 -> DSD1024 is small step

 

If I read this correctly the second quote dealt with DSD to DSD upsampling.

[Le Concombre Masqué]

INSOMNIA tonight ; listened to beloved Leonard Cohen's Ten New Songs

 

24/44 source & ext2/ASDM5EC to 128 output

 

listened probably 1.5 to 2 dB louder than what I had in mind as reference level ; never sounded so gorgeous.

 

Actually I got shivers.

 

BUT

 

in a half an hour time lapse there's been a good  half a dozen cpu clock throttles alerts signalling from 1 to 1471 events

 

Honestly I did not hear anything wrong, on the contrary, I was amazed

 

So, is it because I was ecstatic or shouldn't we worry about cpu load (i keep it  in the 25% ballpark, varying filters with rate) and heat and their negative impacts on SQ as long as it plays? Or is there a threshold (ie more alerts or more events per alert) at which we should question our perception if we keep being ecstatic ?

[sledwards]

The new ASDM7EC sounds wonderful. I am able to run it with poly-sinc-xtr-mp at DSD128 (upsampling Redbook) on my server, which has an Intel i7-8700 processor. Below are cpu process screen shots from htop for DSD128 and DSD256. Total cpu load for DSD128 is about 388% and 497% for DSD256, which noticeably stutters during playback. I am guessing this does correlate with the 100% load that comes and goes on one of the 8 cores. For comparison, Jussi is running his i7-8086K at DSD256 with a total cpu load of 225%. I looked at the Passmark performance scores for the two processors: i7-8700 = 15,156 and the i7-8086K = 16,691. Not a huge difference maybe, but I really don't know the correlation of the scores with respect to HQPlayer performance. So if I really want to run DSD256 with the new modulator (now), I could swap in the 8086K chip at the low price of 500 to 700 dollars! I think this would be premature. Hopefully there is another more cost effective solution that will even handle DSD512. Eager to hear what others are achieving.

 

        DSD128

 

      DSD256

 

[Miska]

i9-9900K could be also good candidate for CPU.

 

With i7-8086K it is likely all about clock speed. All cores end up running at about 4.5 GHz clock. Without GPU's help the 8086K is also on the edge and needs some careful work on settings.

 

Note, as usual, you can try to tune the core allocation in these edge cases by changing "multicore" attribute between "auto" and "1". This corresponds to grayed and checked in Desktop.

 

I've done some optimizations and will probably put out 4.11.1 release with those. The same tunings/optimizations will also land in Desktop and Pro. Not much faster, but can better utilize multicore code path.

 

[sledwards]

Thanks for the suggestion Jussi. I forgot to mention I am running Embedded. I am not familiar with adjusting core allocation. How is this  performed in Embedded?

[rando]

@Miska are we nearing the tipping point away from desktop processors into server hardware?

 

Considering the possibility of Xeon with a Tesla GPU.  The latter having more CUDA cores than a 2080ti is attractive.  

 

[Miska]

7 minutes ago, rando said:

@Miska are we nearing the tipping point away from desktop processors into server hardware?

 

Considering the possibility of Xeon with a Tesla GPU.  The latter having more CUDA cores than a 2080ti is attractive.  

 

GV100 would be interesting GPU, but nearly impossible to get. In Turing, Nvidia again has cut on the double precision floating point capacity.

 

Server CPUs are not very nice because they tend to have lot of cores, but not particularly high clock frequencies. With GPU, critical point is "single thread" performance in a case where all cores are loaded, which could mean that CPU drops close to base clock frequency.

 

Here are clock frequencies as reported by /proc/cpuinfo of the 8086K when playing DSD256 with ASDM7EC:

cpu MHz		: 4402.130
cpu MHz		: 4402.126
cpu MHz		: 4397.454
cpu MHz		: 4354.592
cpu MHz		: 4364.532
cpu MHz		: 4400.641
cpu MHz		: 4402.135
cpu MHz		: 4402.150
cpu MHz		: 4305.493
cpu MHz		: 4343.871
cpu MHz		: 4334.970
cpu MHz		: 4379.682